New chip opens door to AI computing at the speed of light

Engineers at the University of Pennsylvania have developed a new chip that uses light waves rather than electricity to perform the complex mathematics essential to training AI. The chip has the potential to radically accelerate the processing speed of computers while reducing their power consumption.

The silicon-photonics (SiPh) chip design is the first to bring together the pioneering research of Medal winner Benjamin Franklin and Professor H. Nedwill Ramsey Nader Engheta in manipulating materials at the nanoscale to perform mathematical calculations using light – the fastest possible means of communication – with the SiPh platform, which uses silicon, the cheap and abundant element used to mass produce computer chips.

The interaction of light waves with matter represents a possible path to developing computers that go beyond the limitations of today’s chips, which rely on essentially the same principles as chips from the early days of the 1960s computing revolution.

In an article published in Natural photonicsEngheta’s group, along with that of Firooz Aflatouni, associate professor of electrical and systems engineering, describes the development of the new chip.

“We decided to join forces,” says Engheta, taking advantage of the fact that Aflatouni’s research group pioneered nanoscale silicon devices.

Their goal was to develop a platform for performing so-called vector-matrix multiplication, a mathematical operation essential in the development and operation of neural networks, the computer architecture that powers AI tools of today.

Instead of using a silicon wafer of uniform height, Engheta explains, “you make the silicon thinner, say 150 nanometers,” but only in specific regions. These height variations, without the addition of any other materials, provide a means of controlling the propagation of light through the chip, since height variations can be distributed to cause light to scatter in specific patterns, allowing the chip to perform mathematical calculations. at the speed of light.

Due to constraints imposed by the commercial foundry that produced the chips, Aflatouni says, this design is already ready for commercial applications and could potentially be adapted for use in graphics processing units (GPUs), demand for which has exploded with the generalization of chips. interest in the development of new AI systems.

“They can adopt the Silicon Photonics platform as an add-on,” says Aflatouni, “and then you can accelerate training and classification.”

In addition to faster speed and lower power consumption, Engheta and Aflatouni’s chip has privacy benefits: Since many calculations can be done simultaneously, there will be no need to store sensitive information in a computer’s working memory, making a future computer powered by such technology virtually impossible to hack. .

“No one can hack into non-existent memory to access your information,” says Aflatouni.

Other co-authors include Vahid Nikkhah, Ali Pirmoradi, Farshid Ashtiani and Brian Edwards of Penn Engineering.